Evolutionary Search for Automated Design of Uncertainty Quantification Methods

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Seleznyov, Mikhail, Korbut, Daniil, Moskvoretskii, Viktor, Somov, Oleg, Panchenko, Alexander, Tutubalina, Elena
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917383806910464
author Seleznyov, Mikhail
Korbut, Daniil
Moskvoretskii, Viktor
Somov, Oleg
Panchenko, Alexander
Tutubalina, Elena
author_facet Seleznyov, Mikhail
Korbut, Daniil
Moskvoretskii, Viktor
Somov, Oleg
Panchenko, Alexander
Tutubalina, Elena
contents Uncertainty quantification (UQ) methods for large language models are predominantly designed by hand based on domain knowledge and heuristics, limiting their scalability and generality. We apply LLM-powered evolutionary search to automatically discover unsupervised UQ methods represented as Python programs. On the task of atomic claim verification, our evolved methods outperform strong manually-designed baselines, achieving up to 6.7% relative ROC-AUC improvement across 9 datasets while generalizing robustly out-of-distribution. Qualitative analysis reveals that different LLMs employ qualitatively distinct evolutionary strategies: Claude models consistently design high-feature-count linear estimators, while Gpt-oss-120B gravitates toward simpler and more interpretable positional weighting schemes. Surprisingly, only Sonnet 4.5 and Opus 4.5 reliably leverage increased method complexity to improve performance -- Opus 4.6 shows an unexpected regression relative to its predecessor. Overall, our results indicate that LLM-powered evolutionary search is a promising paradigm for automated, interpretable hallucination detector design.
format Preprint
id arxiv_https___arxiv_org_abs_2604_03473
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Evolutionary Search for Automated Design of Uncertainty Quantification Methods
Seleznyov, Mikhail
Korbut, Daniil
Moskvoretskii, Viktor
Somov, Oleg
Panchenko, Alexander
Tutubalina, Elena
Computation and Language
Artificial Intelligence
Uncertainty quantification (UQ) methods for large language models are predominantly designed by hand based on domain knowledge and heuristics, limiting their scalability and generality. We apply LLM-powered evolutionary search to automatically discover unsupervised UQ methods represented as Python programs. On the task of atomic claim verification, our evolved methods outperform strong manually-designed baselines, achieving up to 6.7% relative ROC-AUC improvement across 9 datasets while generalizing robustly out-of-distribution. Qualitative analysis reveals that different LLMs employ qualitatively distinct evolutionary strategies: Claude models consistently design high-feature-count linear estimators, while Gpt-oss-120B gravitates toward simpler and more interpretable positional weighting schemes. Surprisingly, only Sonnet 4.5 and Opus 4.5 reliably leverage increased method complexity to improve performance -- Opus 4.6 shows an unexpected regression relative to its predecessor. Overall, our results indicate that LLM-powered evolutionary search is a promising paradigm for automated, interpretable hallucination detector design.
title Evolutionary Search for Automated Design of Uncertainty Quantification Methods
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2604.03473